Matlab-vision包学习-Image Registration and Geometric Transformation

在这一篇中，我们将介绍一个新的函数：estimateGeometricTransform，其它的函数都在之前介绍学习过了。

函数/Functions

函数名称：estimateGeometricTransform

功能：利用匹配的特征点对估计几何变换(Similarity,affine,projective transformations)

语法：tform = estimateGeometricTransform(matchedPoints1, matchedPoints2,transformType);

[tform,inlierpoints1,inlierpoints2] = estimateGeometricTransform(matchedPoints1, matchedPoints2,transformType);

[_,status] = estimateGeometricTransform(matchedPoints1, matchedPoints2,transformType);

[ _ ] = estimateGeometricTransform(matchedPoints1, matchedPoints2,transformType,Name,Value);

其中，matchedPoints1和matchedPoints2分别表示两个图像匹配的对应点；tform是几何变换对象(affine2d object/projective2d object)；transformType是可以取值为’similarity'/'affine'/'projective'，即几何变换的类型；status为返回状态，具体取值如下表；inlierpoints1和inlierpoints2分别为两个图像对应的内点；Name为用一对单引号包含的字符串，Value为对应Name的值。

status及含义

status	含义
0	无错误
1	matchedPoints1 和 matchedPoints2点不足
2	足够的内点没有发现

Name&Value参数

Name	Value
'MaxNumTrials'	默认值为1000，范围为正整数，为了获取最大内点进行随机试验的次数，取较大的值时，能够提高估计的鲁棒性
'Confidence'	默认值为99，范围为(0,100)，获取最爱内点的置信度，取较大值时能够提高估计的鲁棒性
'MaxDistance'	默认值为1.5，范围为>0，可以衡量误差

举例：

close all;
clear all;
clc;
% 原图像
original = imread('cameraman.tif');
imshow(original);
title('Base image');
% 退化图像
distorted = imresize(original,0.7);
distorted = imrotate(distorted,31);
figure;
imshow(distorted);
title('Transformed image');
% 提取SURF特征点
ptsOriginal = detectSURFFeatures(original);
ptsDistorted = detectSURFFeatures(distorted);
% 获取描述子
[featuresOriginal,validPtsOriginal] = extractFeatures(original,ptsOriginal);
[featuresDistorted,validPtsDistorted] = extractFeatures(distorted,ptsDistorted);
% 描述子匹配
index_pairs = matchFeatures(featuresOriginal,featuresDistorted);

% 匹配的特征点
matchedPtsOriginal = validPtsOriginal(index_pairs(:,1));
matchedPtsDistorted = validPtsDistorted(index_pairs(:,2));

%显示匹配结果
figure;
showMatchedFeatures(original,distorted,matchedPtsOriginal,matchedPtsDistorted);
title('Matched SURF points, including outliers');

%估计几何变换矩阵
[tform,inlierPtsDistorted,inlierPtsOriginal] = estimateGeometricTransform(matchedPtsDistorted,matchedPtsOriginal,'similarity');
figure;
showMatchedFeatures(original,distorted,inlierPtsOriginal,inlierPtsDistorted);
title('Matched inlier points');

%反变换图像
outputView = imref2d(size(original));
Ir = imwarp(distorted,tform,'OutputView',outputView);
figure;
imshow(Ir);
title('Recovered image');